Using Optical Flow Trajectories to Detect Whitecaps in Light-Polluted Videos

Abstract

Whitecap formation is an important factor in the exchange of momentum, heat, and gas on the ocean surface. The long-term measurement of whitecaps is necessary to deepen our understanding of the mechanisms of ocean surface motion. However, traditional detection methods are highly sensitive to illumination. Under various illumination conditions, significant light pollution may be introduced into images. The poor performance caused by using images degraded with light pollution is not conducive to automated long-term whitecap measurement. In this study, we propose a two-step method for the detection of whitecaps under various illumination conditions. An abnormal detection method based on previous whitecap detection methods for the accurate detection of whitecaps in light-polluted areas is proposed as the first step. Using the detection results, we propose a post-processing method based on optical flow trajectories at two sampling rates to separate actual whitecap components in images containing false positives. Experiments show that the method proposed in this study can more accurately detect whitecaps in images with light pollution when compared to existing methods.